Catálogo de publicaciones - libros

The increasing amount of data used for classification, as well as the demand for complex models with a large number of well tuned parameters, naturally lead to the search for efficient approaches making use of massively parallel systems. We describe the parallelization of support vector machine learning for shared memory systems. The support vector machine is a powerful and reliable data mining method. Our learning algorithm relies on a decomposition scheme, which in turn uses a special variable projection method, for solving the quadratic program associated with support vector machine learning. By using hybrid parallel programming, our parallelization approach can be combined with the parallelism of a distributed cross validation routine and parallel parameter optimization methods.

Mobile databases are becoming more available and thus are drawing more attention from both research and industrial communities. They are currently being widely used in devices such as cell phones, hand-held devices, and notebook computers, among others. Broadcasting is a scalable way to send data from a server to multiple clients. Broadcasting algorithms must be constructed in a way that minimizes the average waiting time for clients. XML is a new standard for representing data in a hierarchical structure, and has many advantages over relational representations due to its portability, flexibility, readability, and customizability. XML has recently been deployed onto many mobile devices; thus a new kind of broadcasting algorithm should be constructed to address the unique characteristics of the way XML databases are queried and accessed. In this paper, we presented a new kind of broadcasting algorithm (BA) by utilizing association-rules in clients’ request trends. We implemented three BAs: namely, Exhaustive, Recursive, and Greedy. We tested and compared our BAs with the conventional BAs: namely Sequential and Popularity. The experimental results show that our BAs utilizing association rules perform better than the conventional BAs in both skewed-request situations and requests with association-rules.

Making a decision has often many results and repercussions. These results don’t have the same importance according to the considered phenomenon. This situation can be described by the introduction of the cost concept in the learning process. In this article, we propose a method able to integrate the costs in the automatic learning process. We focus our work on the misclassification cost and we use decision trees as a supervised learning technique. Promising results are obtained using the proposed method.

A comparative study for assessing the reliability of protein-protein interactions in a high-throughput dataset is presented. We use various state-of-the-art classification algorithms to distinguish true interacting protein pairs from noisy data using the empirical knowledge about interacting proteins. Then we compare the performance of classifiers with various criteria. Experimental results show that classification algorithms provide very powerful tools in distinguishing true interacting protein pairs from noisy protein-protein interaction dataset. Furthermore, in the data setting with lots of missing values like protein-protein interaction dataset, K-Nearest Neighborhood and Decision Tree algorithms show best performance among other methods.

Organizations have adopted various data mining techniques to support their decision-making and business processes. However, the mining analysis is not performed and supervised by the final user, the management of the organization, since the knowledge of mathematical models as well as expert database administration skills is required. This paper describes a distributed architecture for association rule mining analysis in the retail area, designed to be used directly by the management of an organization and implemented as a Java web application. The rule discovery algorithm is executed at the database server that hosts the source data warehouse, while the only used client tool is a web browser. The user interactively initiates the rule discovery process through a simple user interface, which is used later to browse, sort and compare the discovered rules.

We have developed a computer-aided detection system for detecting lung nodules, which generally appear as circular areas of high opacity on serial-section CT images. Our method detected the regions of interest (ROIs) using the density values of pixels in CT images and scanning the pixels in 8 directions by using various thresholds. Then to reduce the number of ROIs the amounts of change in their locations based on the upper and the lower slices were examined, and finally a nodule template based algorithm was employed to categorize the ROIs according to their morphologies. To test the system’s efficiency, we applied it to 276 normal and abnormal CT images of 12 patients with 153 nodules. The experimental results showed that using three templates with diameters 8, 14 and 20 pixels, the system achieved 91%, 94% and 95% sensitivities with 0.7, 0.98 and 1.17 false positives per image respectively.

MIGP () realizes information retrieval and integration in extensive distributed medical information systems, which adapts to the essential requirement for the development of healthcare information infrastructure. But the existing MIGPs, which are constructed mostly based on database middleware, are very difficult to guarantee local hospital data security and remote accessing legality. In this paper, a MIGP based on the WSRF-compliant HL7 () grid middleware is proposed, which aims to combine the existing HL7 protocol and grid technology to realize medical data and image retrieval through the communications and interoperations with different hospital information systems. We also design the architecture and bring forward a metadata-based scheduling mechanism for our grid platforms. At last, experimental MIGPs are constructed to evaluate the performance of our method.

In this paper, a structural and event based multimodal video data model (SEBM) is proposed. SEBM supports three different modalities that are visual, auditory and textual modalities for video database systems and it can dissolve these three modalities within a single structure. This dissolving procedure is a mimic of human interpretation regarding video data. The SEBM video data model is used to answer content-based, spatio-temporal and fuzzy queries about video data. A SEBM prototype system is developed to evaluate the practical usage of the SEBM video data model when storing and querying the video data.

The aim of this paper is to investigate the feasibility of predicting the gender of a text document’s author using linguistic evidence. For this purpose, term- and style-based classification techniques are evaluated over a large collection of chat messages. Prediction accuracies up to 84.2% are achieved, illustrating the applicability of these techniques to gender prediction. Moreover, the reverse problem is exploited, and the effect of gender on the writing style is discussed.

The management of modern telecommunications networks must satisfy ever-increasing operational demands. We propose a study for the improvement of intelligent administration techniques in telecommunications networks. This task is achieved by integrating knowledge base of expert system within the management information used to manage a network. For this purpose, an extension of OSI management framework specifications language has been added and investigated. For this goal, we shall use the language Guidelines for the Definition of Managed Objects (GDMO) and a new property named RULE which gathers important aspects of the facts and the knowledge base of the embedded expert system. Networks can be managed easily by using this proposed integration.